The Crash of Flight 447

Air France Flight 447 crashed into the Atlantic Ocean (near here) some time ago. In the absence of much physical evidence, experts figured out that the crash involved a misunderstanding of air speed due in part to faulty air speed data from iced over sensors. Then, last April, the cockpit data and voice recorders were fished out of the sea, which is rather incredible. Since then, further analysis has confirmed the initial finding but added a twist: Yes, there was a problem with air speed, but the proximate reason for the crash is because the individuals piloting the plane (especially one of them) misunderstood, well, how to fly a plane because the things they needed to do were not part of their training.

The voice recording is transcribed and analysed in a new book, in French, , which is in turn reported and discussed here. Riveting.

And I’ll remind you of this post: Flight 447 and the ITCZ, which is more about the ITCZ than the flight or its demise.

Share and Enjoy:
  • Twitter
  • StumbleUpon
  • Facebook
  • Digg
  • Yahoo! Buzz
  • Google Bookmarks
  • LinkedIn

15 thoughts on “The Crash of Flight 447

  1. OK, that clears things a bit. I have been wondering why the icing of the pitot tubes would be a problem, because they had two other independent ways to get their speed: GPS and the power of their engines that generate the said speed. But the autopilot had only the tubes. So it switched off, to make way for the better human judgement … that just was’t there.

    So why? Why wasn’t the autopilot reading GPS or engine thrust? Why hasn’t the airline industry learned anything from the Three Mile Island disaster?

  2. @Lassi:

    You do not know what you are talking about. The airline industry has learned a lot from every disaster they have. They will learn a lot from this one.

    You seem to think that engine thrust or GPS speed is a valid indicator for flying. You couldn’t be further from the truth.

    Actual airspeed, as in the speed you are moving through a section of air is of critical importance. GPS speed is meaningless for flying a plane, as well as how much thrust the engines are putting out, that gives you no indication as to how fast air is going over the wings.

    Here’s an example: You have to maintain 350knots at a certain high altitude and you have a margin of +/- 10knots before the plane goes into a stall. You are flying into a 60knot headwind. Your GPS speed would be 290knots, as that’s how fast you are moving over the ground, but you are actually traveling 350knots compared to the air you are in. If you lose pitot tube readings (these measure actual airspeed), and relied on GPS speed, you’d fall out of the sky just like AF447.

    The true problem with AF447 is that pilots were not trained properly for when the aircraft becomes unable to fly itself, as most Airbus aircraft do everyday. Pilots merely suggest to the aircraft what to do more than 99% of the time with Airbus, and also to a certain extent Boeing aircraft. This is a human interface issue, and will be addressed.

    Fun fact: An Airbus will not allow the pilot to perform an “unsafe” maneuver, such as a rapid pulling up, as that over stresses the wings. A Boeing, however, will let the pliot rip the wings off the plane if they want to. Boeing assumes the pilot knows what’s best for the aircraft.

  3. RE: Why aren’t the pitot tubes equipped with de-icing heaters that prevent them from getting clogged with ice?

    Answer: They are. All IFR certified aircraft (which definitely includes commercial passenger jets) are required to have pitot heaters.

    However, they are not “always on”, are not automatic and have to be turned on manually. (I suspect this because heating the pitot tube potentially changes the accuracy of the instrument itself. Warm air is less dense and could, at least in theory, indicate a different reading than the unheated air.)

    As a private pilot myself, I am certain that the flight crew, at least at some point in their career, *was* taught to recognize and deal with the effects of a blocked pitot system. Unfortunately, it wasn’t recently enough to trigger that thought when things started going badly. My understanding of the events was that due to the type of storm they flew into, the icing was both sudden and very rapid, and that the crew probably had plenty of turbulence to deal with, which drove their focus to maintaining simple control of the aircraft among the bumps, amped up the adrenaline and did not bode well for calm, clear rational thought process. Being night and probably piss poor visability besides, the crew had no outside visual references to rely on, and when things got bad and instrumentation didn’t match what they experienced, they reacted…badly. They didn’t have time to sit back and analyze the data, and made bad decisions about the validity of the different information they were receiving, and how best to maintain control of the aircraft.

    Like most major accidents, this one was caused by chain of events and circumstances that stacked up in the wrong direction. Had any one link been broken, the outcome could have likely been different.

    I can foresee the FAA mandating that pitot heat systems on commercial jet aircraft be engaged automatically whenever air temp / dew point spreads come to within a few degrees, and that if heating a pitot tube does affect its accuracy, that airspeed and altitude indicators be adjusted automatically, or otherwise visually indicate in the PFD that pitot heat is being applied.

  4. Kuz, I get your point about airspeed, but the fact remains that we can now look at the GPS speed indicator retrospectively using the “black box” data and note that for a very long time the GPS speed was something like 60-90 knots. Also, the altimeter was indicating dropping altitude. How is it possible that noting that the GPS speed was slower than takeoff speed would not give a clue that the aircraft was stalling, especially when combined with the stall alarm going off for several minutes?

  5. Fun fact: An Airbus will not allow the pilot to perform an “unsafe” maneuver, such as a rapid pulling up, as that over stresses the wings.

    Seriously? Keeping the stick back and the nose up during a stall for 20 minutes is not unsafe? That is not a fun fact. That is some kind of strange denialism!

  6. David: They actually did turn on de-icing equipment and the air speed indicators were returned to normal status. Then, they flew more or less correctly for several minutes, then accidentally drove the aircraft into the ocean at a 45 degreee angle. More or less.

    They were messed up by the freezing indicators, rattled by the ITCZ, then they flaked out and demonstrated that it is probably not a good idea to have a system whereby pilots are “unable to fly a plane’ that is not flying itself

    After reading the transcript, I came to the conclusion that an experienced Cessna pilot who know where the controls were would have kept that plane flying.

  7. By the way, when the Airbus was first built, there was concern that having a jet liner that would pretty much fly itself would lead to a pattern of training and experience that would cause these planes to crash when the plane could not actually fly itself. I think it is worth acknowledging, now that we know the facts, that this Airbus crashed, taking all on board to their deaths, for that exact reason.

  8. In response to: “Fun fact: An Airbus will not allow the pilot to perform an “unsafe” maneuver, such as a rapid pulling up, as that over stresses the wings. A Boeing, however, will let the pliot rip the wings off the plane if they want to. Boeing assumes the pilot knows what’s best for the aircraft.”

    Well, interestingly, it was an Airbus that had it’s tail snapped by an overzealous pilot handling the rudder, not a Boeing. There should not be a possibility for the pilot to snap the whole vertical stabilizer like that in an airliner.

    I think Boeing is doing more things right in regards to fly-by-wire and controls (yoke vs. sidestick). Airbus should switch over to yokes which move in tandem.

  9. Greg, they did not dive the airplane into the ocean at 45 degrees. The angle of the aircraft’s path (relative to the ground) was 45 degrees, because it was stalled. The aircraft itself was pointing up, about 15 degrees give or take (just like on take-off) while it was descending at 45 degrees towards the ocean, with its nose up. Picture an aircraft taking off, in that position, except it is flying down, not where the nose is pointing.

  10. I was totally talking about the angle of decent. And yes, the nose was pointing up a bit.

    My point was that if you are in an aircraft that is falling at that speed and going that slow you can’t possibly miss the fact if you look at the altimeter and the GPS speed estimate.

    The reason the aircraft crashed is that the pilots did pretty much everything they could do to drive it into the sea, and they did that becuase they did not know how to fly an airplane, and they did not know how to fly an airplane because they were never trained because they were flying an infallible machine.

  11. @Kuz: actually, as an engineer I do know a bit about what I was writing about.

    I have no doubt about the airline industry leaning from its own disasters, but are they learning from disasters in other industries?

    Three Mile Island wasn’t an airplane, it was a nuclear power plant. Yet it should have been a lesson also for airlines. The reactor was lost in very much the same manner: all sorts of alarms were ringing, and the operators couldn’t figure out what was happening. Yet if they had noticed one small gauge they could have saved the reactor. It was the one that indicated level in the overflow tank. It was rising, because that’s where all the coolant was going. So yes, it was a user interface issue, and a very famous one for us engineers.

    The lesson from TMI was to desing the alarm system so that important alarms turn off less important ones. The operators have a better chance of understanding the situation when they are not distracted by secondary events. Another improvement is to have several sensors cross-checked to filter out silly readings, and alert the operators to fix that sensor. In an ideal case, the control computer should be reading all sensors, and compare them to a set of scenarios. If someting goes wrong, it could suggest the most likely scenario for the operators.

    In an Airbus you can get speed reading also from other sources. They are not as good as the pitot tubes – if the pitot tubes are working. AF447 was cruising high above the Atlantic. At those speeds you can make a good guess from available data, e.g. correct GPS reading with wind speed (from weather map) to get air speed. Or use the computer model that predicts air speed from engine thrust, load, attack angle, etc. Those models exist. They are produced in the design phase. Give that speed to the pilot, but mark it as “unreliable”.

    But no, the autopilot just switched off, leaving the pilots in limbo. The airline industry seems to have learned nothing from the TMI disaster.

  12. The lesson from TMI was to desing the alarm system so that important alarms turn off less important ones. The operators have a better chance of understanding the situation when they are not distracted by secondary events.

    Yeah, but hold on a sec here! The alarm that was sounding of flight 447 was the stall alarm. The aircraft was in a stall. That’s why it crashed. The user interface worked fine, but the pilots ignore it!

    Personally, I think every air plane should have at least one pilot who has flown a real airplane before.

  13. That was such a depressing read. When it was announced long ago that engineers suspected the air speed indicator was iced and not working I was telling my buddies that still doesn’t make sense – landing without the air speed indicator is a bitch but level flight is trivial. Basically the pilot in command stalled the aircraft and kept it in a stall. If the idiot had only let go of the yoke the aircraft would have gently adjusted itself and maintained altitude. It reads like the sort of jokes you’d hear in training: “I don’t understand why we’re stalled, I’m pulling the nose up as far as I can.” The more senior first officer wasn’t any better; he was trusting the idiot to do the right thing (and the idiot continued to pull back on the yoke despite being instructed multiple times to descend, which should be interpreted as ‘push the yoke forward ever so slightly’). If he had only checked the attitude indicators he should have realized that someone was pulling back on the yoke. I didn’t know Airbus had yokes which acted independently. At least on a Boeing you know someone’s fiddling with the control and you can scream at them. The article also states that the Airbus averages the input from the 2 yokes – I wonder if that’s true and why the system would ever be designed that way.

Leave a Reply

Your email address will not be published.